STUDY DEMOCRACY AND HUMAN RIGHTS UNEQUAL DEMOCRACIES: WHO DOES (NOT) VOTE? Voter turnout trends in the OSCE region since 1970 Democracy is the promise of political equality for citizens expressed in free and fair elections. When more and more people no longer vote, democracy suffers. Low voter turnout means socially unequal voter turnout. People with lower levels of education and members of professions with lower social status are disproportionally less likely to vote when voter turnout drops. This effect of socially unequal voting is especially prominent among young voters. A translation of the German or i ginal »Wer geht(nicht) wählen?« by Lea Elsässer, Armin Schäfer and Jonas Wenker June 2022 The smallest turnout differences is found in the egalitarian democracies of Northern Europe whereas stronger differences subsist in the young Eastern European democracies and Anglo-Saxon countries.

FES DEMOCRACY OF THE FUTURE – WHO DOES(NOT) VOTE? DEMOCRACY AND HUMAN RIGHTS UNEQUAL DEMOCRACIES: WHO DOES (NOT) VOTE? Voter turnout trends in the OSCE region since 1970 B

 Content 1 INTRODUCTION  2  VOTER TURNOUT AND DIFFERENCES IN TURNOUT MATTER  3  AND METHODS 4 RESULTS   6 5 THE LAW OF DISPERSION   8  PARTICIPATES? THE INTERPLAY BETWEEN DIFFERENT SOCIAL VARIABLES   10 7 CONCLUSION  14  15  17 Data 18

FES DEMOCRACY OF THE FUTURE – WHO DOES(NOT) VOTE? 1 INTRODUCTION Political participation in a democracy goes far above and beyond voting. Voting in elections remains the most fundamental way of expressing one’s political concerns and participating in shaping the political community, however. No other form of participation is used by as many citizens and no other political event attracts as much attention as national parliamentary or presidential elections. Even minimalist theories of democracy therefore view free, equal and fair elections as a basic prerequisite for classifying a country as democratic. Consequently, it is also important how many people participate in elections and whether and to what extent we observe participation gaps between social groups. Mass non-voting cannot be interpreted – as it indeed has been in the past – as an expression of satisfaction, but instead points to a problem afflicting democracy. In this report, we examine voter turnout levels and differences in turnout in more than 170 elections in 29 countries that are part of the Organisation for Security and Co-operation in Europe(OSCE) over time. We show that advocates of political equality should not be indifferent when it comes to low turnout. Not only does voter turnout differ significantly between countries and regions, but also between social groups. Strong differences in participation are evident both between social classes and educational groups, especially in the younger age groups. In the following section, we first explain why the level of voter participation is important in normative terms. We then present the data and the methodological approach of our analysis. In the empirical section of this study, we then explore cross-national trends. In the conclusion, we summarise the main findings of the study. 2

Why voter turnout and differences in turnout matter 2 WHY VOTER TURNOUT AND DIFFERENCES IN TURNOUT MATTER Unequal participation is the unsolved problem of democracy, as the political scientist Arend Lijphart(1997) noted as early as the end of the 1990s. In addition to the normative obligation to achieve at least legal equality, democracy also embodies the promise that all social groups and their concerns will be heard equally. If factors exogenous to democratic procedures systematically influence the(non-)participation of certain social groups, however, the democratic promise of equality runs the danger of being sullied and debased. Various reasons can be singled out as to why unequal participation in elections is problematic. Firstly, voting in elections serves as an opportunity to evaluate the work of an incumbent government and – on the basis of this evaluation – to bring about a change in government. Voting thus fundamentally communicates how the population assesses the work of the past government and in what political direction the citizens would like to see things move in the future. If only part of society participates in this vote on the future, communication between the government and the governed is skewed(Verba 2003: 666). This is especially the case when the group of voters systematically differs in its political views from the group of non-voters. The social composition of the group of non-voters is therefore of great importance because it provides information on whether the political perspectives of certain social groups are systematically communicated less. sources available to parties to mobilise potential voters are scarce, decisions must be made on how to use these resources. It may appear rational to focus on groups or neighbourhoods that display a higher voter turnout(Roßteutscher/ Schäfer 2016; Hajnal/ Trounstine 2005). Moreover, there is clear evidence that such feedback also impacts policy decisions that are made. Those who do not vote(or articulate their own interests otherwise) may count less than politically active voters when contentious decisions are at stake(Lijphart 1997: 3–4). Various studies show that the preferences of poorer citizens, which(increasingly) account for the majority of non-voters, are often disregarded in policy decisions(Elsässer 2018; Elsässer et al. 2021). Thus, in countries with higher voter turnout – all other things being equal – not only social spending(Hicks/ Swank 1992; Crepaz 1998) is higher, but also the degree of redistribution (Mahler 2008), which results in income inequality being less pronounced there(Mueller/ Stratmann 2003; Anderson/ Beramendi 2008; Chong/ Olivera 2008; Mahler 2010). 1 Pontusson and Rueda(2010) produce similar findings, showing that centre-left parties in countries with high voter turnout position themselves further to the left than in countries with low voter turnout. Although the decision not to participate appears at first glance to be a private decision, these non-randomly distributed decisions cumulate into political impact. Along with this, the election result itself could be influenced by who participates in the election. The extent to which the level of voter participation changes the election result is a subject of controversy in political science(cf. among others Radcliff 1994; Pacek/ Radcliff 1995; Fisher 2007; Rosema 2007; Kohler/ Rose 2010; Schäfer 2012). But here as well, one can say: the more citizens differ from one another, the greater the probability that there will be differences in preferences on the one hand and that these will be reflected in voting behaviour on the other. Thirdly, in addition to distorted communication on the part of the population, there can be interrelationships between differences in voter turnout and the behaviour of political parties. With regard to political mobilisation of the population – be it to vote in elections or to engage in party politics – persistent differences in turnout can lead to a vicious cycle between low turnout and lack of mobilisation. When the reAgainst this background, this study aims to examine the differences in voter turnout for as many countries from the OSCE area as possible from the 1970s onwards. The focus is on differences between education groups and social classes (captured by occupational groups). Education and occupation are not only suitable indicators to capture socio-economic position in society, but people’s position in the labour market and the accompanying daily experiences at the work place also shape their political preferences(Erikson/ Goldthorpe 1992: 31; Kitschelt 1994; Manza/ Brooks 2010). Hence, systematic differences in participation go hand in hand with the problems described above. 1 In the USA, there is also evidence that regions and social groups with low voter turnout are disadvantaged in the allocation of government funds(Hill/ Leighley 1992; Hill et al. 1995; Martin 2003; Hajnal 2010: ch. 5). 3

FES DEMOCRACY OF THE FUTURE – WHO DOES(NOT) VOTE? 3 DATA AND METHODS In order to investigate social inequality in electoral participation, existing social science data sets were identified and harmonised to enable a comparative analysis of electoral participation by social groups over time and across as many OSCE countries as possible. In the first step, the Varieties of Democracy(V-DEM) data was used to determine which OSCE countries can be classified as democracies according to common definitions(see figure 10 in the appendix). To this end, average scores from the Liberal Democracy Index, which is frequently used in democracy research, were recorded for the years since 1990. Of the 57 member states, 34 countries have scores above 0.5 during this period, which we assume to be the threshold level for liberal democracy. A total of 20 countries from the group of OSCE states are consistently governed democratically. In order to analyse elections in the democratic OSCE states, existing data was compiled from various social science surveys that capture both the socio-economic status of the respondents and their participation in respective national parliamentary or presidential elections. National election studies and cross-country comparative studies were used. National studies of individual states usually cover a longer period of time, but are only available for a few of the countries we examined or only partially allow comparison with present-day data due to technical or methodological changes. In sum, five studies could be identified that allow an evaluation of longer time series. Good data coverage up to the 1970s or 1980s is thus available for the USA(American National Election Study, ANES), the United Kingdom(British Election Study, BES), Switzerland(Swiss Election Study, SES), the Netherlands(Dutch Parliamentary Election Study, DPES) and Germany(General Population Survey of the Social Sciences, ALLBUS). Cross-national studies, on the other hand, make it possible to comparatively analyse a large number of states. We use the Comparative Study of Electoral Systems(CSES) and the European Social Survey(ESS) for this purpose. 2 These two datasets cover a large share of the democratic OSCE countries, although the first wave of the CSES only begins in 2 Other large survey projects such as the European Values Study, the World Values Survey or the International Social Survey Programme do not enquire about participation in the last election or do so only very irregularly. 1996 and the earliest elections examined in the ESS date from 1999. The combination of these datasets allows for a relatively comprehensive coverage of the democratic OSCE countries for the period from 2000 to 2019. 3 Besides basic availability, the quality of the data is of crucial importance in the analyses to be carried out. In order to avoid bias in the analyses, minimum standards were set that had to be met in order to include the respective country in the analysis. For example, a response rate of 80 per cent was specified for questions key to the analysis and applied in the selection of the data sets. 4 With the help of the described procedure, a data set based on over 700,000 respondents was created, covering 176 elections in 29 OSCE states. Figure 1 shows that until around the turn of the millennium there was little coverage, but thereafter good coverage, of these countries. After merging the different data sets, they were harmonised, as different questions and measuring instruments were used in some cases. The focus of the study is on educational groups and social classes, which had to be categorised and harmonised accordingly. In addition, the age and gender of respondents were also included and recorded to allow comparison across the data sets. The harmonised education variable allows a comparison of persons without a secondary school qualification(level 1), with a secondary school qualification(level 2) and with a qualification higher than secondary school(level 3). The social class of the respondents was harmonised using the simplified class scheme according to Oesch(2006). This class scheme takes into account changed occupational structures in post-industrial societies, since the»working class« includes not only manual industrial workers, but also employees in service professions whose occupations do not 3 An overview of all data sets used can be found in the»data sources section« in the appendix. 4 The CSES data was only used when no other data was available for the respective election. There are two reasons for this: First, the CSES data is often generated from existing national election studies, for example, the CSES data for the USA comes from ANES. However, simultaneous use of the same data by two different datasets can lead to bias. In addition, CSES data sometimes appear to be of lower quality than data from alternative sources. Higher-quality sources were therefore preferred wherever possible. 4

Data and methods Figure 1 Elections included in the analysis Belgium – Bulgaria – Denmark – Germany – Estonia – Finland – France – Greece – United Kingdom – Ireland – Iceland – Italy – Canada – Croatia – Latvia – Lithuania – Netherlands – Norway – Austria – Poland – Portugal – Sweden – Switzerland – Slovakia – Slovenia – Spain – Czech Republic – Hungary – United States – 1970 1980 – – – – – – 1990 2000 2010 2020 require an academic degree(e. g. cooks, parcel delivery workers, preschool teachers). The simple scheme distinguishes five social classes: 1. Higher-grade service classes, which include employees in highly qualified professions as well as higher liberal professions(lawyers, physicians, etc.) and large business-owners; 2. Lower-grade service classes ( e. g. skilled administrative staff, mechanical technicians and social workers); 3. Small business-owners; 4. Skilled workers (e. g. office workers, salespersons and electricians); and 5. Unskilled and semi-skilled workers(e. g. security guards, waiters and parcel delivery workers). 5 5 One precondition for classification using the class scheme is that the classification of respondents in the International Standard Classification of Occupation according to the scheme of 1988 or 2008 is available. For the detailed 16-level class scheme according to Oesch, three variables each are required for respondents and their life partners; these are only available for very few elections, however. For this reason, the five-level scheme was opted for and modified so that it can be mapped based on two variables for the respondents. This allows comparative analysis over a much longer period of time. 5

FES DEMOCRACY OF THE FUTURE – WHO DOES(NOT) VOTE? 4 RESULTS In the following sections we present the results of the data analysis and focus in particular on inequality in voter turnout between social groups. First, however, it is worth looking at how the overall turnout has changed over time in the countries we studied. 6 Empirical electoral research has identified a number of factors that can explain individual participation in elections and the level of voter turnout (cf. Jackman/ Miller 1995; Blais 2006). At the individual lev6 The»country-specific diagrams« section in the appendix contains diagrams on the individual countries. el, these are, first of all, resource endowment(education, income, wealth and time) and, secondly, basic attitudes such as political interest, acceptance of elections as a norm, identification with a political party, but also the feeling of being able to make a difference, as well as satisfaction with the functioning of democracy(cf. Cancela/ Geys 2016). At the macro level, electoral laws(e. g. whether registration is automatic, whether voting is compulsory or how proportional the electoral system is), election-specific factors such as the contested nature of the election outcome, and socio-economic factors such as income distribution are important(cf. Geys 2006). Figure 2 Change in voter turnout over time 20% Czech Republic Slovakia Portugal Slovenia Croatia Bulgaria Latvia Lithuania 10% Deviation 0% –10% –20% 1970 1980 1990 Note: The deviation from the national average for all elections under consideration is shown for each election. Data: www.idea.int/vt. 6 Slovakia 2000 Croatia United States Hungary Greece Portugal 2010 2020

Results Figure 3 Level of voter turnout in the last national parliamentary election Belgium Sweden Denmark Iceland Netherlands Norway Germany Austria France Italy Spain United States Hungary Finland United Kingdom SlovakiaCzech Republic Estonia Ireland Canada Poland Greece Latvia Slovenia Bulgaria Portugal Lithuania Croatia Switzerland Data: www.idea.int/vt. 40% 50% 60% 70% 80% 90% 100% While factors explaining turnout levels are well researched, it is less clear why turnout is falling in many developed democracies(Gray/ Caul 2000; Blais et al. 2004). In particular, existing findings raise one question: If institutional barriers to voting have tended to be reduced and the average resource endowment of citizens has increased at the same time, why is voter turnout falling in many countries? Recent studies indicate that it is not so much absolute resource endowment, but rather its distribution that is decisive with regard to voter turnout, especially among poorer groups: the greater the social inequality, the lower and more unequal the voter turnout(Mahler 2002; Solt 2008, 2010; Scervini/ Segatti 2012; Schäfer 2015; Schäfer/ Schwander 2019). Figure 2 shows the historical evolution of turnout in democratic elections in the countries under study. The zero line stands for the average turnout for each country, while the dots show the extent to which a specific election deviates from this average. Positive scores indicate above-average turnout, negative scores below-average turnout. 7 Three things can be inferred from Figure 2: First, there are significant deviations from the averages, indicating considerable changes in turnout over time. Secondly, we see major positive deviations in the early 1990s, which can be attributed to the new democracies in Central and Eastern Europe. In the first free elections, turnout levels in these countries were often very high, but the subsequent decline was all the more precipitous. Thirdly, there are many more negative deviations from the national average in the second half of the period under observation. This is partly, but not entirely, explained by the point just mentioned. Voter turnout has also declined in many places in the Western European countries, however. Only a few of the countries we examined still achieve voter turnout levels above 80 per cent today(figure 3). Apart from Belgium, where voting is compulsory, only Sweden, Denmark and Iceland do so. Voter turnout is highest in the egalitarian Scandinavian countries and the Netherlands. Most countries have turnout rates between 60 and 80 per cent, while the rate is below this level in eight countries. Overall, many Eastern European countries have low turnout rates. The lowest turnout is in Switzerland, however, where the combination of a proportional representation government and a highly developed direct democracy pushes turnout to an unusually low level. 7 The development of voter turnout for all individiual 29 countries can be accessed at https://democracy.fes.de/topics/inequality-democracy. In the next step, we turn to an examination of how the level of voter turnout is related to differences in turnout. 7

FES DEMOCRACY OF THE FUTURE – WHO DOES(NOT) VOTE? 5 THE LAW OF DISPERSION As far back as the 1930s, the Swedish scientist Herbert Tingsten(1975) investigated how the level of voter turnout related to differences in participation. After compiling and analysing extensive data sets, he developed the»law of dispersion«: the lower the voter turnout, the greater the social differences in voter participation between resource-poor and resource-rich citizens(Tingsten 1975: 230). Several later studies have tended to confirm this correlation, prompting Kohler(2006: 170171) to posit:»Low voter turnout is generally unequal voter turnout«(cf. also Gallego 2015; Armingeon/ Schädel 2015). So if democracy requires not only legal equality, but also approximately equal participation, low voter turnout dashes the promise of political equality(Verba 2003; Schäfer 2015). Can it still be confirmed today – more than 80 years after Tingsten’s original study – that low turnout goes hand in hand with greater differences in participation? To test this, in Figure 4 we plot on the horizontal axis the level of turnout for all the elections under study and on the vertical axis the level of differences in turnout between the lowest and the highest occupational group. Where overall turnout is high, as we can see from the figure, participation gaps are small. As turnout decreases, however, differences between occupational groups increase. The correlation is not perfect, but it is systematic, so it is often possible to infer where inequality in turnout is higher or lower by comparing turnout rates. Figure 4 General turnout and differences in turnout between the lowest and highest occupational groups 40 Differences in voter turnout in % 30 20 10 50% 60% 70% 80% Voter turnout 8 90% 100%

The law of dispersion Figure 5 Comparative national voter turnout and turnout differences Iceland – Denmark – Belgium – Sweden – Norway – Greece – Italy – Germany – Spain – Finland – Netherlands – Austria – Hungary – Ireland – Bulgaria – France – Slovenia – Slovakia – Portugal – Croatia – United Kingdom – Poland – Estonia – Latvia – Switzerland – Lithuania – Czech Republic – 40% 60% Working class Note: The statistically predicted probability of voting is shown taking into account country, age and gender. 80% Upper service class 100% Estimated voter turnout rates The extent of the differences can be seen more precisely when we estimate the level of turnout for unskilled and semi-skilled workers on the one hand and for members of the higher-grade service class on the other. In order to exclude the effect of other variables, we estimate a statistical model that uses class as the primary explanatory variable, but also controls for age and gender. From the interaction between occupational group and country, we can estimate the turnout rates for the two occupational groups in each country, adjusted in this manner. The results are shown in Figure 5. The bright red points indicate estimated turnout by members of the higher-grade service classes, while the turquoise points show that of employed workers. The grey line between the dots indicates the size of the turnout gap. On the one hand, the graph confirms the picture that participation differences are greater in countries with lower average turnout than in countries with high turnout. In Iceland, the country with the highest reported turnout, even in the lower occupational group 90 per cent of respondents state that they voted. This figure rises to 98 per cent in the upper occupational groups. In comparison, in Hungary there are 18 percentage points between workers and higher-grade service classes, and in Switzerland, Lithuania and the Czech Republic – three countries with particularly low turnout – even more than 25 percentage points. Secondly, the graph shows that there are systematic differences between different regions: While the majority of the countries with the largest participation differences are Eastern European countries, the differences are smallest in Northern European countries and Belgium. The realisation that the law of dispersion still applies today is not trivial, as theoretically low turnout could also be evenly distributed across classes. As the analyses show, however, this is usually not the case. It is mainly citizens with fewer resources who stay away from the ballot box. Since the data for the countries analysed cover different time periods, it is not possible to show average trends over time without distorting the analysis. For this reason, the differences in voter turnout between different social groups(in addition to occupational group, education, age and gender are also taken into account) over time separately for each country are not shown here. They can instead be accessed at: https://democracy.fes.de/topics/inequality-democracy. 9

FES DEMOCRACY OF THE FUTURE – WHO DOES(NOT) VOTE? 6 WHO PARTICIPATES? THE INTERPLAY BETWEEN DIFFERENT SOCIAL VARIABLES Having noted that voter turnout is falling in many countries and that low voter turnout is associated with high socio-economic inequality, we now take a more nuanced look at how voter participation differs between different social groups. We devote particular attention to the interaction between social class(or education) and gender, or social class and age. To ensure a high data quality, in this section we only use information on elections that have taken place since 2000. In Figure 6, persons surveyed are differentiated according to their social class as well as gender. The height of the bars shows the average level of turnout across all countries in the five occupational groups, with the level of turnout for men and women being shown separately within each occupational group. As can be seen from the figure, occupational groups differ more in electoral participation than women and men. While workers participate significantly less than lower and higher-grade service classes, the participation of female and male respondents within the respective classes is very similar. A different picture emerges when we examine class differences within different age groups. Figure 7 shows the voter turnout for four age groups broken down by social class. Three points are worth mentioning here: First, electoral participation is higher on average among the over-60 age group than among younger age groups. Second, there are clear turnout differences between higher and lower-grade occupational groups in each age group. And thirdly, these Figure 6 Differences in voter turnout by gender and social class 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% m m än al n e lich w fe e m ib a li le ch m m ä a n l n e lich w fe e m ib a li le ch m m än al n e lich w fe e m ib a li le ch U ob p e p r e e r D se ie r n vi s c t e kl c a l s a s s e s Lo m w i e t r t le g re ad D e ie s n er s v tk ic l e as c s l e ass K S l m ei a n l g l b ew us e in rb e e s t s r e o i w be n n e d rs e Note: The representation is based on all elections since 2000. m m ä a n l n e lich w fe e m ib a l l i e ch Fa S c k h i a ll r e b d ei w te o r r _ k in e n rs en m m än al n e lich w fe e m ib a li le ch Un U n & sk A il n le g d el a e n rn d te semi-skilled workers 10

Who participates? The interplay between different social variables Figure 7 Differences in voter turnout by age and social class 1001%100 1 0% 0 1%0 0 0% % 0% 95%9955%9%5 5 9% % 5% obeoroeb U boeD p erbioer p eebnrD e eDs r irteiDek s en e linDsae r tssni v ktesskl i enta c lksas e tlsakes c sle l as a es s s s e mittml L mei o rtmiet w ltimeDtl e trile r teet nrDl g eDsi r teiD a eken d linDsae e tssniktessk s lenta e lksas r tlsa v kessl i ea c ses e se class KleiKnKlgelKeinlwiKeng S ilegne m rewigbnw a eeg l wter l erbr b weberi u ebtrt s reerb i nte n ierbdiebteeri s enbe s ndeibd o neeed w ned n e ers FachFaFarcFbhc S aheaF k ciarh i tbr l cea l bhe e re_iab d tirentbe w irnte_re_i o ntrnie_n r n k irnne_ e enin r ne s nnen U U n n -U& s Un k nUAil -n& l U e &-gnA d &eA-nl a &enAgr n gnenA d elgtenleer s glnr e ent m rletener i tn ete skilled workers 90%9900% 9 %0 0 9% % 0% 85%8855%8%5 5 8% % 5% 80%8800% 8 %0 0 8% % 0% 75%7755%7 7 %5 5 7% % 5% 70%7700%7 7 %0 0 7% % 0% 65%6655%6 6 %5 5 6% % 5% 60%6600% 6 %0 0 6% % 0% 55%555%5 5 %5 5 5% % 5% 50%5500% 5 %0 0 5% % 0% unt u eu n urnn d 3utet0n e eurt r rne3 3 3tr0e0 0 3r030 Note: The diagram is based on all elections since 2000. 30– 3 43340 0 0–3 – –40 4 43–40 4 4–444 45– 4 64405 5 5–4 – –65 6 640–0 0 56–060 übeür o üb6 v büe0 e erbü r re6b 6 6r0e0 0 6r060 Figure 8 Differences in voter turnout by age and education 100% 100% 95% 95% 90% 90% 85% 85% 80% 80% 75% 75% 70% 70% 65% 65% 60% 60% 55% 55% 50% 50% u u n n d te e r r 3 3 0 0 Note: The diagram is based on all elections since 2000. 3 3 0 0 – 4 4 4 4 n l i o e w drig m m it e t d el ium h h o i c g h h 4 4 5 5 – –60 0 o ü v b e r r 6 6 0 0 11

FES DEMOCRACY OF THE FUTURE – WHO DOES(NOT) VOTE? Voter turnout in % Figure 9 Differences in voter turnout by age and education in five regions ANGLO-SAXON COUNTRIES 100 80 60 40 20 0 under 30 100 30–44 45–60 CENTRAL EUROPEAN COUNTRIES 80 60 40 20 0 under 30 100 30–44 SCANDINAVIA 45–60 80 60 40 20 0 under 30 100 30–44 EASTERN EUROPE 45–60 80 60 40 20 0 under 30 100 30–44 SOUTHERN EUROPE 45–60 80 60 40 20 0 under 30 Note: The diagram is based on all elections since 2000. 30–44 Education: high middle 45–60 low 12 over 60 over 60 over 60 over 60 over 60

Who participates? The interplay between different social variables differences are particularly pronounced among younger respondents. Especially among the under-30s, there is a clear participation gap between higher and lower-grade service classes on the one hand and employed workers on the other. If one assumes – as research on the habitual component of voting behaviour suggests(Plutzer 2002; Gerber et al. 2003; Fowler 2006) – that the decision to vote or not to vote is conditioned by socialisation at a younger age and changes only gradually in the further course of life, differences in voter participation could increase further in coming decades. When we focus on education groups instead of occupational groups(Figure 8), a very similar picture emerges. Among those under 30, differences in voter turnout are particularly large at 20 percentage points, while the difference for the strata over 60 years of age is only eleven points. Not all young people stay away from the ballot box, however. Rather, this is especially the case for those with lower education levels(Roßteutscher / Abendschön 2014; Schäfer et al. 2020). In the last step of the analysis, we examine differences in voter turnout by age and education in a regional comparison. As shown above, voter turnout is particularly high in Northern Europe, while it is lower in the(Southern) Eastern European countries. In line with this picture, differences in participation in the Scandinavian countries are smaller than in other regions. In Central Europe(Belgium, Germany, the Netherlands, Austria and Switzerland) the differences are also moderate, whereas in the Southern European countries (France, Greece, Italy, Portugal and Spain) they are at a medium level. Particularly large differences – especially among younger people – are to be found in the(South) Eastern European countries(Bulgaria, Estonia, Croatia, Latvia, Lithuania, Poland, Hungary, Slovakia, Slovenia) and the Anglo-Saxon countries(Great Britain, Ireland, Canada and the USA). Not only is the sharpest decline in voter turnout to be observed in the eastern EU Member States, but also increasing social inequality in voter participation, which is particularly pronounced among the under-30 age group. 13

FES DEMOCRACY OF THE FUTURE – WHO DOES(NOT) VOTE? 5 CONCLUSION This study investigated trends in voter turnout in 29 OSCE countries and patterns of inequality that can be identified along different social indicators. It was based on survey data from 29 countries, some of which date back to the 1970s. For most countries, however, reliable survey data has only been available since the turn of the millennium, so that we were often only able to look at the shorter period. The following points summarize the key takeaways: First, voter turnout is falling in many countries, but wide disparities across countries remain. In the Scandinavian countries, voter turnout remains high, whereas in many Eastern European countries it has fallen to low levels by international standards. Secondly, low voter turnout implies high social inequality in voter turnout. This»law of dispersion«, which was articulated as far back as in the 1930s, still applies today and has been once again corroborated by our analysis. When average voter turnout drops, this indicates a disproportionately strong decline in voting among social groups that have fewer resources. This social inequality in voter participation becomes apparent when educational groups or social classes are compared. Citizens with lower education or those from the working class are much more likely to stay away from the ballot box on election day than those with higher formal education or members of the higher and lower-grade service classes. Thirdly, participation differences along class lines are particularly pronounced among younger age groups: In no age group are differences between higher and lower social classes greater than among those under 30. In contrast, only minor differences in voter participation can be observed between women and men. Although elections continue to be the pivotal element when it comes to democratic participation, social groups make unequal use of the opportunity to make their voices heard through the ballot box. Unequal participation remains the »unsolved problem« of democracy(Lijphart 1997), which is based on the promise of political equality for all citizens. 14

References REFERENCES Anderson, Christopher J./ Beramendi, Pablo(2008):»Income, ­Inequality, and Electoral Participation«, in: Beramendi, Pablo/ Anderson, Christopher J.(Hrsg.): Democracy, Inequality, and R­ epresentation. A Comparative Perspective, Russell Sage Foundation, New York, S. 278–311. Armingeon, Klaus/ Schädel, Lisa(2015):»Social Inequality in Political Participation: The Dark Sides of Individualisation«, in: West European Politics 38(1), S. 1–27. Bernstein, Robert/ Chadha, Anita/ Montjoy, Robert(2001):»Overreporting Voting. Why it Happens and Why it Matters«, in: Public ­Opinion Quarterly 65, S. 22–44. Blais, André(2006):»What Affects Voter Turnout«, in: American Review of Political Science 9, S. 111–125. Blais, André et al.(2004):»Where Does Turnout Decline Come From?«, in: European Journal of Political Research 43, S. 221–236. Cancela, João/ Geys, Benny(2016):»Explaining Voter Turnout: A Meta-Analysis of National and Subnational Elections«, in: Electoral Studies 42, S. 264–275. Chong, Alberto/ Olivera, Mauricio(2008):»Does Compulsory Voting Help Equalize Incomes?«, in: Economics& Politics 20(3), S. 391–415. Coppedge, Michael et al.(2020): V-Dem Dataset 2020, https://doi. org/10.23696/vdemds20. Crepaz, Markus M. L.(1998):»Inclusion versus Exclusion: Political Institutions and Welfare Expenditures«, in: Comparative Politics 31(1), S. 61– 80. Elsässer, Lea(2018): Wessen Stimme zählt?, Campus Verlag, Frankfurt/ New York. Elsässer, Lea/ Hense, Svenja/ Schäfer, Armin(2021):»Not just money: Unequal responsiveness in egalitarian democracies«, in: Journal of European Public Policy 28(12), S. 1890–1908. Erikson, Robert/ Goldthorpe, John H.(1992): The Constant Flux. A Study of Class Mobility in Industrial Societies, Clarendon Press, Oxford. Fisher, Stephen D.(2007):»(Change in) Turnout and(Change in) the Left Share of the Vote«, in: Electoral Studies 26, S. 598–611. Fowler, James H.(2006):»Habitual Voting and Behavioral Turnout«, in: Journal of Politics 68(2), S. 335–344. Gallego, Aina(2015): Unequal Political Participation Worldwide, Cambridge University Press, New York. Gerber, Alan S./ Green, Donald P./ Shachar, Ron(2003):»Voting May Be Habit-Forming: Evidence from a Randomized Field Experiment«, in: American Journal of Political Science 47(3), S. 540–550, https://doi.org/­10.1111/1540-5907.00038. Geys, Benny(2006):»Explaining Voter Turnout: A Review of Aggregate-Level Research«, in: Electoral Studies 25(4), S. 637–663. Gray, Mark/ Caul, Miki(2000):»Declining Voter Turnout in Advanced Industrial Democracies, 1950 to 1997. The Effects of Declining Group Mobilization«, in: Comparative Political Studies 33(9), S. 1091–1122. Hajnal, Zoltan L.(2010): America’s Uneven Democracy. Race, Turnout, and Representation in City Politics, Cambridge University Press, ­Cambridge. Hajnal, Zoltan/ Trounstine, Jessica(2005):»Where Turnout Matters: The Consequences of Uneven Turnout in City Politics«, in: Journal of ­Politics 67(2), S. 515–535. Hicks, Alexander M./ Swank, Duane H.(1992):»Politics, Institutions, and Welfare Spending in Industrialized Democracies 1960–82«, in: American Political Science Review 86(3), S. 658–674. Hill, Kim Quaile/ Leighley, Jan E.(1992):»The Policy Consequences of Class Bias in State Electorates«, in: American Journal of Political Science 36(2), S. 351–365. Hill, Kim Quaile/ Leighley, Jan E./ Hinton-Andersson, Angela (1995):»Lower-Class Mobilization and Policy Linkage in the U.S. States«, in: American Journal of Political Science 39(1), S. 75–86. Jackman, Robert W./ Miller, Ross A.(1995):»Voter Turnout in the Industrial Democracies During the 1980s«, in: Comparative Political Studies 27(4), S. 467–492. Kitschelt, Herbert(1994): The Transformation of European Social ­Democracy, Cambridge University Press, Cambridge. Kohler, Ulrich(2006):»Die soziale Ungleichheit der Wahlabstinenz in Europa«, in: Alber, Jens/ Merkel, Wolfgang(Hrsg.): Europas Osterweiterung: Das Ende der Vertiefung?, WZB-Jahrbuch 2005, Edition Sigma, ­Berlin, S. 159–179. Kohler, Ulrich/ Rose, Richard(2010):»Under What Circumstances Could Maximising Turnout Alter an Election Result?«, in: Representation 46(2), S. 119–138. Lijphart, Arend(1997):»Unequal Participation: Democracy’s Unresolved Dilemma«, in: American Political Science Review 91(1), S. 1–14. Mahler, Vincent A.(2002):»Exploring the Subnational Dimension of Income Inequality: An Analysis of the Relationship Between Inequality and Electoral Turnout in the Developed Countries«, in: International Studies Quarterly 46(1), S. 117–142. Mahler, Vincent A.(2008):»Electoral Turnout and Income Redistribution by the State: A Cross-National Analysis of the Developed Democracies«, in: European Journal of Political Research 47, S. 161–183. Mahler, Vincent A.(2010):»Government Inequality Reduction in Comparative Perspective: A Cross-National Study of the Developed World«, in: Polity 42(4), S. 511–541. Manza, Jeff/ Brooks, Clem(2010):»Class and Politics«, in: Lareau, ­Annette/ Conley, Dalton(Hrsg.): Social Class. How Does it Work?, Russell Sage Foundation, New York, S. 201–231. Martin, Paul S.(2003):»Voting’s Rewards: Voter Turnout, Attentive ­Publics, and Congressional Allocation of Federal Money«, in: American Journal of Political Science 47(1), S. 110–127. Mueller, Dennis C./ Stratmann, Thomas(2003):»The Economic ­Effects of Democratic Participation«, in: Journal of Public Economics 87, S. 2129–2155. Oesch, Daniel(2006): Redrawing the Class Map: Stratification and institutions in Britain, Germany, Sweden and Switzerland, Palgrave Macmillan Houndmills, Basingstoke/ Hampshire/ New York. Pacek, Alexander/ Radcliff, Benjamin(1995):»Turnout and the Vote for Left-of-Centre Parties: A Cross-National Analysis«, in: British Journal of Political Science 25, S. 137–143. Plutzer, Eric(2002):»Becoming a Habitual Voter: Inertia, Resources, and Growth in Young Adulthood«, in: American Political Science Review 96(1), S. 41–56. Pontusson, Jonas/ Rueda, David(2010):»The Politics of Inequality: Voter Mobilization and Left Parties in Advanced Industrial States«, in: Comparative Political Studies 43(6), S. 675–705. Radcliff, Benjamin(1994):»Turnout and the Democratic Vote«, in: American Politics Quarterly 22, S. 259–276. Rosema, Martin(2007):»Low Turnout: Threat to Democracy or Blessing in Disguise? Consequences of Citizens’ Varying Tendencies to Vote«, in: Electoral Studies 26, S. 612–623. 15

FES DEMOCRACY OF THE FUTURE – WHO DOES(NOT) VOTE? Roßteutscher, Sigrid/ Abendschön, Simone(2016):»Wahlbeteiligung junger Erwachsener. Steigt die soziale und politische Ungleichheit?«, in: Roßteutscher, Sigrid/ Rosar, Ulrich/ Faas, Thorsten(Hrsg.): Bürger und Wähler im Wandel der Zeit. 25 Jahre Wahl- und Einstellungsforschung in Deutschland, Springer VS, Wiesbaden. Roßteutscher, Sigrid/ Schäfer, Armin(2016):»Asymmetrische Mobilisierung: Wahlkampf und ungleiche Wahlbeteiligung«, in: Politische Vierteljahresschrift 57(3), S. 455–483. Scervini, Francesco/ Segatti, Paolo(2012):»Education, Inequality and Electoral Participation«, in: Research in Social Stratification and Mobility 30(4), S. 403–413. Schäfer, Armin(2012):»Beeinflusst die sinkende Wahlbeteiligung das Wahlergebnis? Eine Analyse kleinräumiger Wahldaten in deutschen Großstädten«, in: Politische Vierteljahresschrift 53(2), S. 240–264. Schäfer, Armin(2015): Der Verlust politischer Gleichheit: Warum die sinkende Wahlbeteiligung der Demokratie schadet, Campus-Verlag, Frankfurt am Main. Schäfer, Armin/ Roßteutscher, Sigrid/ Abendschön, Simone (2020):»Rising start-up costs of voting: political inequality among firsttime voters«, in: West European Politics 43(4), S. 819–844. Schäfer, Armin/ Schwander, Hanna(2019):»›Don’t play if you can’t win‹: does economic inequality undermine political equality?«, in: European Political Science Review 11(3), S. 395–413. Solt, Frederick(2008):»Economic Inequality and Democratic Political Engagement«, in: American Journal of Political Science 52(1), S. 48–60. Solt, Frederick(2010):»Does Economic Inequality Depress Electoral Participation? Testing the Schattschneider Hypothesis«, in: Political Behavior 32, S. 285–301. Tingsten, Herbert(1975): Political Behavior. Studies in Election Statistics, Arno Press, London. Verba, Sidney(2003):»Would the Dream of Political Equality Turn out to Be a Nightmare?«, in: Perspectives on Politics 1(4), S. 663–679. 16

Appendix APPENDIX Country selection Since we only analyse elections in liberal democracies of the OSCE region, we first used the V-DEM dataset(Coppedge et al. 2020) to determine the level of the Liberal Democracy Index for each country on average since 1990. Only countries above the threshold of 0.5 were included in the analysis. Since we the use the average as a baseline, some countries are currently scoring higher or lower on the Index. We intended to use transparent criteria for our country selection. Analysing voter turnout in non-democratic countries seems not reasonable. Our classification is not overly restrictive. For instance, Hungary and Poland today no longer classify as liberal democracies but are part of the analysis. See Lührmann, Anna; Lindberg, Staffan I.(2019): A third wave of autocratization is here: what is new about it? In: Democratization 26(7), S. 1095–111. The average scores for each OSCE country on the Liberal Democracy Index are shown in the figure below.  Figure 10 Democracy score for OSCE countries Sweden – Denmark – Norway – Germany – Portugal – Finland – Switzerland – Estonia – Belgium – Ireland – Spain – Netherlands – Greece – United States – United Kingdom – France – Iceland – Czech Republic – Luxembourg – Austria – Italy – Canada – Poland – Slovenia – Lithuania – Slovakia – Cyprus – Latvia – Hungary – Croatia – Malta – Bulgaria – Mongolia – Romania – Serbia – Moldova – Albania – Bosnia and Herzegovina – Georgia – North Macedonia – Turkey – Montenegro – Ukraine – Armenia – Kyrgyzstan – Russia – Kazakhstan – Belarus – Tajikistan – Azerbaijan – Uzbekistan – Turkmenistan – 0,0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 Data: V-DEM Project 17

FES DEMOCRACY OF THE FUTURE – WHO DOES(NOT) VOTE? DATA SOURCES Aarts, Kees/ Todosijevic, Bojan/ van der Kaap, Harry(1971–2006): Dutch Parliamentary Election Study Cumulative Dataset. Alt, James/ Crewe, Ivor/ Sarlvik, Bo(1974): British Election Study, Cross-Section Survey. American National Election Studies(1948–2016): ANES Time Series Cumulative Data File, ICPSR – Interuniversity Consortium for Political and Social Research. American National Election Studies(2021): ANES Time Series Study, ICPSR – Interuniversity Consortium for Political and Social Research. Baur, Robert et al.(1971–2019): Swiss Election Study(Selects), cumulative dataset, FORS – Swiss Centre of Expertise in the Social Sciences. Center for Political Studies, University of Michigan, Ann Arbor (2020): CSES Integrated Module Dataset(IMD): CSES – Comparative Study of Electoral Systems. Clarke, Harold et al.(2005): British Election Study, Face-to-Face Survey. European Social Survey(2018): European Social Survey(ESS), Cumulative Data Wizard, NSD – Norwegian Centre for Research Data. Fieldhouse, Ed et al.(2015): British Election Study, Face-to-Face Post-Election Survey, UK Data Service. Fieldhouse, Ed et al.(2019): 2019 BES Post-Election Random Probability Survey, v.1.0.0. GESIS – Leibniz-Institut für Sozialwissenschaften(1980–2018): Allgemeine Bevölkerungsumfrage der Sozialwissenschaften ALLBUS, GESIS Data Archive. Heath, Antony/ Jowell, Roger/ Curtice, John(1983): British General Election Study, Cross-Section Survey. Heath, Antony/ Jowell, Roger/ Curtice, John(1987): British General Election Study, Cross-Section Survey, 2nd Edition. Heath, Antony et al.(1992): British General Election Study, Cross-Section Survey. Heath, Antony et al.(1997): British General Election Study, Cross-Section Survey, 2nd Edition. Robertson, David/ Crewe, Ivor/ Sarlvik, Bo(1974): British Election Study. Robertson, David/ Sarlvik, Bo/ Crewe, Ivor(1979): British Election Study, Cross-Section Survey. The Comparative Study of Electoral Systems(2021): CSES Module 5, Advance Release. Whiteley, Paul/ Sanders, David(2010): British Election Study, Face-toFace Survey. 18

Imprint ABOUT THE AUTHORS IMPRINT Lea Elsässer is a post-doctoral researcher at the Institute for Political Science at Johannes-Gutenberg University Mainz. Her research focuses on the effects of social inequality on political representation. In her dissertation, she investigated the effects of social inequality and political responsiveness in Germany. Armin Schäfer is a professor of comparative politics at Johannes-Gutenberg University Mainz. He is an expert of empirical democracy research and specifically interested in the relation between social and political equality. Jonas Wenker is a PhD candidate at the Institute for Political Science at Johannes-Gutenberg University Mainz. He researches the development of democratic attitudes with a special focus on supporters of populist parties. Publisher: FES Regional Office for International Cooperation – Democracy of the Future Reichsratsstr. 13/5 A-1010 Vienna Responsibility for content: Johanna Lutz, Director, Democracy of the Future Project Management: Michael Jennewein, Researcher for Democracy and Economics Project website: democracy.fes.de Contact/Orders: democracy.vienna@fes.de Design: pertext, Berlin| www.pertext.de The views expressed in this publication are not necessarily those of the Friedrich-Ebert-Stiftung or of the organization for which the author works. Commercial use of media published by the Friedrich-Ebert-Stiftung(FES) is not permitted without the written consent of the FES Publications by the FriedrichEbert-Stiftung may not be used for electioneering purposes. ISBN 978-3-98628-157-1 © 2022 www.fes.de/bibliothek/fes-publikationen 21

UNEQUAL DEMOCRACIES: WHO DOES(NOT) VOTE? Voter turnout trends in the OSCE region since 1970 This study examines how voter turnout and differences in voter turnout have trended in 29 OSCE countries since the 1980s. In order to be able to conduct our analysis, we primarily make use of official election data and secondly compile survey data on more than 170 elections in a database. We found that voter turnout tends to be lower today than it used to be in many countries. It has fallen particularly sharply in Eastern European countries. With the help of the survey data, it can also be shown that low voter turnout is a socially unequal phenomenon. Among non-voters are a disproportionate share of people with low formal education and from occupational groups with lower social status. These differences are particularly marked among younger people today. On the other hand, we find little difference in participation of women and men in elections. Differences in participation are smallest in the egalitarian democracies of Northern Europe, whereas they are more pronounced in the young democracies of Eastern Europe and the liberal Anglo-Saxon countries. Further information on the topic can be found here: democracy.fes.de